artashes karapetyan norges bank bogdan stacescu bi … · artashes karapetyan norges bank bogdan...
TRANSCRIPT
This file was downloaded from BI Brage, the institutional repository (open access) at BI Norwegian Business School
http://brage.bibsys.no/bi
Does information sharing reduce the role of collateral as a
screening device?
Artashes Karapetyan Norges Bank
Bogdan Stacescu BI Norwegian Business School
This is the authors’ accepted, refereed and final manuscript to the article published in
Journal of Banking and Finance, 43(2014): 48-57
DOI: 10.1016/j.jbankfin.2014.02.010
Publisher’s version available at http://dx.doi.org/10.1016/j.jbankfin.2014.02.010
Copyright policy of Elsevier, the publisher of this journal:
The author retains the right to post the accepted author manuscript on open web sites operated by author or author's institution for scholarly purposes, when there is no institutional open
access policy or mandate.
http://www.elsevier.com/journal-authors/author-rights-and-responsibilities#author-posting
Does information sharing reduce the role of collateral as a screeningdevice?
Artashes Karapetyana,1,∗, Bogdan Stacescub,∗∗
aNorges Bank, Financial Stability Research, Bankplassen 2, P.O. Box 1179 Sentrum, Norway. Tel: + 47 22 31 62 52bBI Norwegian Business School, Nydalsveien 37, 0484 Oslo. Tel: +47 46 41 05 19
Abstract
Information sharing and collateral are both devices that help banks reduce the cost of adverse selection.
We examine whether they are likely to be used as substitutes (information sharing reduces the need for
collateral) or complements. We show that information sharing via a credit bureaus and registers may
increase, rather than decrease, the role of collateral: it can be required in loans to high-risk borrowers
in cases when it is not in the absence of information sharing. Higher adverse selection makes the use
of collateral more likely both with and without information sharing. Our results are in line with recent
empirical evidence.
Keywords: Bank competition, information sharing, collateral
JEL classification numbers: G21, L13
∗Principal corresponding author∗∗Corresponding author
Email addresses: [email protected] (Artashes Karapetyan), [email protected] (BogdanStacescu)
Preprint submitted to Elsevier July 22, 2013
1. Introduction
Adverse selection is an important issue facing banks (Stiglitz and Weiss [43]). Not all borrowers and
projects applying for bank loans should be funded; however, since banks do not have the same information
as their applicants, deciding which of them are creditworthy can be difficult. In this paper, we look at
two instruments that banks can use to select their borrowers: collateral requirements and credit records.
Collateral can be used to reduce adverse selection since high-quality borrowers are more likely to
pledge assets and thus signal their creditworthiness. This is a well-established result in the theoretical
literature (Bester, [9]; Chan and Kanatas [18]; Besanko and Thakor [8]). Collateral requirements are
widespread in practice (Avery, Bostic, and Samolyk [1]) and have a long history (Bodenhorn, [11]).
Empirical studies have found that there is an inverse relationship between collateral and interest rates
(Berger, Espinosa-Vega, Frame and Miller [5]; Berger, Frame and Ioannidou [4]; Cerqueiro, Ongena and
Roszbach [16]), and that collateral does indeed seem to be used to select borrowers ex ante (Jimenez,
Salas and Saurina [32]; Berger, Espinosa-Vega, Frame and Miller [5]; Berger, Frame and Ioannidou [4]).
Collateral requirements therefore have an important role in credit allocation.
Another useful tool for reducing adverse selection is the information acquired during the lending
relationship (Boot [12]). Borrowers’ performance over successive loans, for instance, can be used to update
the bank’s assessment of their value as a client. Low-quality borrowers will gradually be eliminated from
the pool of loan applicants.
Some of the information acquired during lending relationships, such as repayment histories, is made
available to competing banks through credit bureaus and credit registers (Djankov, McLiesh and Shleifer
[22]; Miller [37]). As a result, banks can use the data received from other lenders to select their potential
borrowers (Jappelli and Pagano [31]). The role and geographical spread of information sharing arrange-
ments have significantly increased in recent years. In a survey of Latin American banks Miller Miller [37]
reports that 93 percent of the banks used credit information for their commercial loans (84 percent did
so for consumer loans and 100 percent for mortgage loans).
Both the information received through credit bureaus and registers and the more traditional collateral
can be used to select loan applicants. Indeed, while credit bureaus are a more recent development in most
countries (Djankov, McLiesh and Shleifer [22]) the frequency of their use in lending decisions has become
comparable. Miller Miller [37] finds that, while collateral still remains important in granting loans,
most bank managers consider payment history as the number one important factor in credit decisions.
Information sharing seems to be associated with better credit allocation (Houston, Lin, Lin and Ma [29]).
Due to potential liquidation costs (Gorton and Kahn [25]; Chen [19]; Benmelech and Bergman [2])
and fluctuations in market value (Bernanke and Gertler [6]; Cerqueiro, Ongena, and Roszbach [16]),
the use of collateral can be expensive for banks and borrowers. The availability of additional data via
information sharing arrangements may provide a potentially cheaper alternative for borrower selection.
When collateral is costly, banks may prefer to reduce the amount required, while still attracting high-
quality borrowers. As a result, it may be interesting to check whether the increasing use of shared credit
2
records is likely to reduce the incidence of collateral requirements.
We analyze the use of collateral and credit records in lending decisions. We find that the overall
picture is quite different from a simple substitution story. Indeed, we show that information sharing
may lead to the use of collateral in circumstances where it would not be required in the absence of a
credit bureau. The reason is that information sharing allows banks to distinguish between borrowers
with different credit histories. Some borrowers will have a good record, but others will have a poor one.
We show that under information sharing there may be a higher incidence of collateral as a result of its
concentrated use for borrowers with bad credit histories.
We build a two-period model with two banks competing for high- and low-quality borrowers. The
banks compete in interest rates and may use collateral to select loan applications. The use of collateral
obviously improves the average borrower quality, but is also costly because of liquidation costs. As in
Gehrig and Stenbacka Gehrig and Stenbacka [24], borrowers face switching costs when moving from one
bank to another.
Borrowers’ history of successful repayments or default also provides information about their credit-
worthiness. Under information sharing, these credit histories become available to the bank that has not
had a lending relationship with a particular borrower.
Whether information is shared or not, liquidation costs imply that collateral will only be used if
adverse selection is important enough. However, information sharing does have an important effect on
the use of collateral, since it allows outside banks to distinguish between pools of borrowers of different
quality.
In the absence of information sharing, banks faced with unknown borrowers can choose to require
collateral, and face liquidation costs in case of default. In the presence of a credit bureau, banks faced
with outside borrowers can distinguish between those with a good credit history, and those with bad
credit events on their record. We show that borrowers with a bad credit history are more likely to be
faced with collateral requirements than they would be in the absence of information sharing. As a result,
the introduction of a credit bureau or a credit register may increase the observed incidence of collateral
requirements.
Our theoretical results are consistent with and provide a theoretical explanation for the empirical
results in Doblas-Madrid and Minetti Doblas-Madrid and Minetti [21]. Using contract-level data from
a U.S. credit bureau, they find that information sharing does not reduce the incidence of collateral, and
that the incidence actually increases for low-quality borrowers.
We also find that, under both information regimes, higher adverse selection makes the use of collateral
more likely. Moreover, higher adverse selection also creates incentives for banks to share information and
make selective use of collateral, and the two work together - if information sharing is not feasible, then
the likelihood of collateral use is lower.
Our study analyzes the bank’s choice of instruments to reduce adverse selection. This is an area that
has received relatively little attention in the literature. An important exception is Manove, Padilla, and
3
Pagano Manove, Padilla, and Pagano [36], showing that the availability of collateral may reduce banks’
screening incentives. We look at another pair of selection instruments, collateral and credit histories,
and also use the idea of cost minimization. The importance of this criterion has been confirmed by the
empirical literature: “the evidence suggests that collateral pledging decisions are generally consistent
with borrowing cost minimization” (Booth and Booth [13]).
While the use of collateral induced by information sharing may increase welfare in our model, we show
that the surplus accruing to high-quality borrowers may actually decrease. This welfare tradeoff is not
necessarily a desirable feature (Gehrig and Stenbacka [24]). It can also be noted that, in addition to the
liquidation costs we model directly in our paper, lenders face the costs of monitoring the pledged assets
(Cerqueiro, Ongena and Roszbach [16]). Moreover, borrowers’ credit availability may change along with
the value of the pledged assets (Gan [23]). When this value is correlated across borrowers, this can amplify
the procyclicality of access to credit (Bernanke and Gertler [6]; Holmstrom and Tirole [28]; Kiyotaki and
Moore [35]). From this angle, our results could be seen as worrying. We find that information sharing and
collateral can be complements: information sharing may increase the likelihood of collateral requirements,
and that increase will actually occur for borrower groups faced with higher adverse selection issues. Thus,
while both information sharing and collateral are tools that help lending decisions, their mixing may lead
to undesirable effects.
We focus on ex ante adverse selection issues rather than ex post moral hazard problems in lending.
Both collateral (Chan and Thakor [18]; Boot and Thakor [12]; Rajan and Winton [41]; Berger, Frame,
and Ioannidou [4]) and information sharing (Padilla and Pagano [38]; Padilla and Pagano [39]) can be
used to reduce moral hazard in lending. Banks’ choice between the two as ex post instruments may be
an interesting issue for further research.
The closest paper to ours in the area of information sharing is Gehrig and Stenbacka Gehrig and
Stenbacka [24]. Looking at a potential downside of credit bureaus, they show that information shar-
ing reduces the returns from establishing banking relationships, and thus weakens competition for the
formation of banking relationships. The result may be higher interest rates for young firms without an
established credit record. In our paper, we identify another potential pitfall of information sharing: the
increase in costly collateral requirements for borrowers faced with significant adverse selection issues.
The rest of the paper is organized as follows. Section II describes the model. Section III and IV solve
for the equilibria under information sharing and in the absence of it, respectively. Section V concludes.
2. The Model
We model the two-period competition between two banks, A and B. They compete for loan contracts
with borrowers who live for two periods, period 1 and 2. Banks raise (unlimited) capital at a fixed cost
r0 per dollar in both periods. In each period they offer a one-period loan contract.
Borrowers form a continuum of length 1. Each of them requires one unit of capital to start a project.
Since they have no funding of their own, they have to borrow the capital from one of the banks. There are
4
two types of borrowers, high (H) and low (L). High-type borrowers have access to a project that returns
a verifiable amount R with probability p and 0 otherwise. Low-type borrowers have a zero probability
of success, but they derive a non-verifiable amount of utility from the business equal to c. In addition,
all borrowers have assets in place of amount C > c, that can be pledged as collateral. The proportion
of H−type (L−type) borrowers is λ (1 − λ) and is common knowledge. We assume that borrowers can
only borrow from one of the banks in each period, but they can borrow from one bank in the first period
and from the other in the second period. We call the bank a given customer has borrowed from during
the first period the “inside” bank for that borrower; the other bank is the “outside” bank.
Before contracting in the first period, banks and borrowers have no information about any of the
borrowers’ type1. The first period thus represents the initial stage of a business, where entrepreneurs
still have much to learn about their business abilities and banks also have relatively little information to
draw on when examining them.
By the end of the first period, however, the inside bank has learned its borrowers’ true type2. Borrowers
also discover their type during the first-period lending relationship. This information is private and
relationship-specific and cannot be communicated credibly to the other bank.
At the end of the first period banks also observe their borrowers’ repayment history. This history can
be made available to the competitor bank under an information sharing arrangement3.
The outside bank can require collateral in the second period to separate high- and low-type borrowers.
Collateral has a proportional liquidation cost L, so that inside banks will never want to require positive
collateral from the high-type borrowers they lend to4. Since an amount of collateral c is enough to screen
borrowers, it is clear that the poaching outside bank will never impose a higher collateral. As borrowers
do not know their type at that point, banks will never require collateral in the first period.
At the beginning of the second period each bank j announces the interest rate rj2 for its high-type first-
period borrowers. (It does not lend to low-type borrowers, since they are obviously not creditworthy.) It
also announces the interest rate ij2 that it offers to the first-period customers of the other bank, as well as
possible collateral requirements for them (cj2). Under information sharing, the successful and defaulting
first-period customers of the competing bank can be offered different contracts.
When borrowers switch to the outside bank, they may bear certain costs due to the cessation of the
relationship with the inside bank. Following Gehrig and Stenbacka [24], we model this as an idiosyncratic
1The assumption that borrowers do not initially know their own type is similar to Manove, Padilla, and Pagano Manove,Padilla, and Pagano [36]. The implication is that very young firms are less likely to pledge collateral, a fact which isconfirmed empirically by Avery, Bostic and Samolyk Avery et al. [1].
2This learning captures the feature of the relationship bank whereby the (incumbent) bank gradually builds its knowledgeof the borrowers’ abilities and their projects. We assume that this learning is perfect. See Sharpe Sharpe [42], Gehrig andStenbacka Gehrig and Stenbacka [24] for similar models of bank learning.
3Gehrig and Stenbacka Gehrig and Stenbacka [24] assume all relationship-specific information can be shared. While ourmodel is robust to this change, we assume that only verifiable hard information can be shared, i.e. the repayment history.See Petersen Petersen [40] for a detailed discussion on hard and soft information.
4It has been shown that the use of collateral is less likely as bank relationships get older and banks presumably learnmore about their customers (Berger and Udell [3]; Bodenhorn [11]; Chakraborty and Hu [17]; Bharath, Dahiya and Saunders[10]). Our assumption captures this stylized fact in a simple way.
5
switching cost that is distributed uniformly on the interval [0, S]. The switching costs may, for example,
reflect the costs of another application procedure at a competitor bank, or the financial costs of transfer-
ring funds from the previous bank. Moreover, as Gehrig and Stenbacka [24] argue, switching costs can
vary largely across customers (see Shy [44]; Kim et al [34], Stango [45]). The switching costs are private
information of the borrowers and are revealed to the borrower at the beginning of period 2. The initial
choice of bank is therefore independent of the switching cost (Gehrig and Stenbacka [24]). As in Sharpe
Sharpe [42], Padilla and Pagano Padilla and Pagano [38], Gehrig and Stenbacka Gehrig and Stenbacka
[24], we assume that successful borrowers will consume their first-period revenues at the end of period 1.
In sum, at the beginning of the first period lenders announce interest rates for a risky borrower
population. At the beginning of the first period, banks may share information and they announce
interest rates and collateral requirements for their existing and potential borrowers. When moving from
the inside to the outside bank at that point, borrowers are faced with switching costs.
The timeline of the bank competition can be summarized as follows:
First period:
1. Banks decide whether or not to share default information.
2. Banks offer interest rates and borrowers choose a bank to borrow from. At the end of the period,
borrowers repay if they can.
Second period:
1. Banks share default information if they have decided to do so.
2. Banks announce inside and outside interest rates and collateral. Borrowers may switch banks if
they are better off doing so.
3. Borrowers repay whenever they can. Pledged collateral is seized if the borrower defaults.
The timeline is illustrated in figure 1.
INSERT FIGURE 1 HERE
We analyze the resulting equilibrium below, first without and then with information sharing.
3. No information sharing
We start by examining the case where banks have decided not to share information. We derive the
subgame perfect Nash Equilibrium via backward induction.
3.1. Equilibrium in the second period
In the second period, banks will try to retain their own first-period high-type borrowers and to “poach”
those of the competing bank.
Let 0 ≤ µi ≤ 1 denote the market share of bank i, i = A,B, acquired in the first period. The inside
bank has acquired information about its first-period borrowers’ true types. It only lends to high-type
borrowers in the second period, and its expected return per loan is pri2 − r0.
6
The outside bank could use the collateral requirement (cj2) to distinguish among the borrowers it tries
to poach. A high-type borrower may switch from bank i to bank j if the switching cost s ∈ [0, S] and
the poaching rate ij2 are low enough: pij2 + (1− p)cj2 + s < pri2. The marginal borrower’s switching cost
is s∗ = pri2− pij2− (1− p)cj2. High-type borrowers that have switching costs above s∗ will stay with their
first-period bank, while the rest will switch.
Alternatively, the outside bank can decide not to require collateral. In that case, it will lend to both
high- and low-type borrowers with relatively low switching costs. The switching cost of the marginal
borrower will be s∗ = pri2 − pij2.
The inside bank’s second-period profits under no information sharing are given by
µiλ(pri2 − r0)1
S
∫ S
s∗ds
where where s∗ will depend on the outside bank’s choice to use collateral or not.
If bank j tries to poach without collateral, then its poaching profits are given by:
µiλ(pij2 − r0)1
S
∫ pri2−pij2
0
ds− roµi(1− λ)
The first term in the equation represents the profits earned on the high-type entrepreneurs successfully
poached from rival bank i. The second term represents the losses from lending to low-type poached
borrowers. Bank i cannot separate these borrowers without collateral.
The losses from lending to low-type borrowers can be avoided if the poaching bank j imposes collateral.
In this latter case its profits are given by:
µiλ(pij2 + (1− p)cj2 − r0 − (1− p)l) 1
S
∫ pri2−pij2−(1−p)cj2
0
ds
where Lcj2 ≡ l represents the liquidation cost of collateral. In order to avoid unnecessary liquidation
costs, the outside bank will set collateral requirements to the minimal level cj2 = c.
Summing up, in the absence of information sharing the total second-period profit of bank i is equal
to
Πi2 = µiλ(pri2 − r0)
1
S
∫ S
pri2−pij2
ds+ (1− µi)λ(pii2 − r0)1
S
∫ pri2−pij2
0
ds− ro(1− µi)(1− λ)
when collateral is not required, and
Πi,c2 = µiλ(pri2 − r0)
1
S
∫ S
pri2−pij2−(1−p)cj2
ds+ (1− µi)λ(pii2 + (1− p)ci2(1− L)− r0)1
S
∫ pri2−pij2−(1−p)cj2
0
ds
7
when collateral is required from the first-period customers of bank j.
Banks will choose interest rates and collateral requirements to maximize their second-period earnings.
The resulting equilibrium is described in the following propositions.
Lemma 1. If collateral is required from outside borrowers, the equilibrium interest rate offered to inside
and outside borrowers are given by r2 = 13p
(2S + 3r0 + (1− p)l
)and i2 = 1
3p
(S + 3r0 + (1− p)(2l− 3c)
),
respectively. Without collateral, the interest rates are given by r2 = 13p
(2S + 3r0
)and i2 = 1
3p
(S + 3r0
).
High-type borrowers with switching costs below s∗ = S3 (s∗ = S
3 −2(1−p)l
3 if collateral is required) switch
to the outside bank.
Proof. See Appendix.
We can see from the expressions above that whenever S > 2(1− p)l (i.e., ri2 > pij2 + (1− p)c) there is
positive switching under contracts with collateral. (There is less switching in the presence of collateral
requirements.) In what follows we will assume that this relationship holds.
Theorem 2. If collateral is required from outside borrowers, the total second-period profits of bank i,
i = A,B, are given by Πi,c2 = µiλ
19S
(2S + (1 − p)l
)2+ (1 − µi)λ 1
9S
(S − (1 − p)l
)2. If collateral is not
required, then total second-period profits are Πi2 = µiλ
49S + (1− µi)
(λ 1
9S − (1− λ)r0).
Proof. See Appendix.
It can be seen from the proposition that, if liquidation costs are high, the banks’ poaching profits can
be lower when poaching is achieved with collateral. The decision concerning collateral requirements is
presented in the following proposition.
Theorem 3. In the absence of information sharing banks will prefer to impose collateral whenever cost
of adverse selection is high enough: r0(1−λ)λ > 2S(1−p)l−(1−p)2l2
9S .
Proof. The proof follows immediately from comparing the expressions for profits from poaching with
and without collateral.
Intuitively, if the adverse selection costs are lower than the total expected liquidation costs of collateral,
then banks will prefer to offer contracts without collateral and with higher interest rates. As we will see
in the next section, information sharing may reduce total liquidation costs, and collateral may then be
preferred by lenders.
8
3.2. Equilibrium in the first period
At the beginning of the first period, banks are faced with a population that consists of both high- and
low-type borrowers. The two types cannot be separated by requiring collateral, since borrowers initially
do not know their own type. Banks will charge an interest rate r1 that maximizes their intertemporal
(two-period) profits (Πi = Πi1 +Πi
2, where Πi1 = µi(λpr
i1− r0), and i = A,B; we assume a discount factor
of 1 between the two periods).
Theorem 4. In the subgame perfect equilibrium first-period interest rates are given by ri,c1 = 1λp
(r0 −
λ 49S
(S+ (1− p)l
)2+λ 1
9
(S− (1− p)l
)2)if collateral is used in the second period and ri1 = 1
λp
(r0−
(λS3 +
(1− λ)r0))
if it is not.
Proof. See Appendix.
The two banks will charge the same first-period interest rate and share the market equally, whether
they anticipate they are going to use collateral later on or not (µA = µB).
Theorem 5. Banks’ intertemporal total profits are given by λ 19
(S − (1 − p)l
)2if collateral is required,
and by λS9 − (1− λ)r0 if it is not.
Proof. The proof follows immediately from plugging the first-period interest rate into the expression
for total intertemporal profits.
Due to the existing of switching costs, bank are always able to secure positive intertemporal rents.
While all profits coming from incumbency rents in the second period are dissipated in the first period
competition for customers, positive profits are secured due to the existence of the poached customers.
This result is similar to that in Gehrig and Stenbacka (2007).
4. Information sharing
We now present the equilibrium interest rates, profits and criterion for using collateral in the case of
information sharing.
In the presence of a credit bureau, lenders commit to reveal the identity of their first-period defaulting
borrowers to competitors. As a result, outside banks can use switching customers’ previous repayment
history in their lending decisions. Borrowers that have successfully repaid their first-period loan are
obviously high-type borrowers; defaulting borrowers are a mix of high- and low-type borrowers. We
analyze below the two possible cases: where the defaulting borrowers are not creditworthy on average,
and where they are.
9
4.1. Bank competition when defaulting borrowers are not creditworthy
Suppose first that defaulting borrowers are not creditworthy (that is, λ(1−p)pλ(1−p)+1−λR − r0 < 0). In
this case, the outside bank’s options are either to require collateral from defaulting, switching borrowers,
or not to lend to them at all. At the same time, successful borrowers are obviously high-type, and the
outside bank will lend to them without collateral. The inside bank will only lend to high-type borrowers,
regardless of their credit histories.
The profits of the inside bank i are given by
λpµi(pri2,N − r0)
1
S
∫ S
s∗N
+λ(1− p)µi(pR− r0)1
S
∫ S
s∗D
where the first term represents profits on successful borrowers (which are offered the competitive rate
ri2,N ) and the second term represents profits on defaulting, but high-type borrowers (who are offered the
monopoly rate R since the outside bank does not try to poach them).
Suppose first that the outside bank j does not bid for the other bank’s first-period defaulting cus-
tomers. It will only bid for successful borrowers, who are obviously high-type. As a result, its profits
from poached customers will be given by:
λpµi(pij2,N − r0)
1
S
∫ s∗N
0
This represents the expected profit earned on customers who have succeeded. It differs from the
respective expression in the absence of information sharing in two ways: the adverse selection cost coming
from lending to low-type borrowers is absent, and the amount of creditworthy, high-type borrowers is
now reduced (by a factor of p).
Suppose next that collateral is imposed by the outside bank for defaulting, switching borrowers.
Because low-type borrowers prefer not to borrow, the outside bank will get only the high-type borrowers
among them. Poaching profits earned on contracts with collateral are thus given by
λpµi(pij2,N − r0)
1
S
∫ s∗N
0
ds+ λ(1− p)(1− µi)(pij2,D + (1− p)cj2(1− l)− r0)1
S
∫ s∗D
0
ds
where the first term represents profits earned on customers who have previously repaid successfully:
the outside bank does not impose collateral on these customers as they are high type by virtue of their
repayment history. The second term represents expected profits earned on customers who are high ability
but who have defaulted, and are now required to post collateral.
Theorem 6. Second-period profits under the contract with and without collateral are given by, respec-
tively, Πi,c2 = µiλpS
49 + µi
λ(1−p)9S (2S + (1 − p)l)2 + (1 − µi)
λpS9 + (1 − µi)
λ(1−p)9S (S − (1 − p)l)2, and
Πi2 = µiλ(1− p)(pR − r0) + µiλp
49S + (1− µi)λp9 S, where i = A,B. High-type borrowers with switching
costs below s∗ = S3 − frac2(1− p)l3 switch to the outside bank if collateral-based loans are available.
10
First-period interest rates are given by ri,c1 = 1λp
(r0− λS3 + 2λ(1−p)2l
3
)and ri1 = 1
λp
(r0− λpS3 − λ(1−
p)(pR− r0))
, respectively.
Two-period (intertemporal) profits are λpS9 + λ(1−p)
9S (S − (1 − p)l)2 when collateral is used to poach
defaulting borrowers, and λp9 S when it is not.
Proof. See Appendix.
By comparing the two possible expressions for second-period earnings one can see that banks will
always prefer to require collateral in the presence of a credit bureau. This was not the case in the absence
of information sharing. The reason is that in the latter case liquidation costs of collateral may outweigh
those of the adverse selection for the poaching bank. However, because previous default information
sharing has revealed type specific information about borrowers, the costs of collateral are reduced by
imposing it only on the high risk (bad history) population. Thus, rather than substituting, collateral
may complement the role of information sharing.
4.2. Bank competition when defaulting creditors are creditworthy
In the previous subsection we have assumed that default information is informative enough to identify
a group of uncreditworthy borrowers. We now look at the case where unsuccessful borrowers are still
creditworthy on average. (In that case there are enough defaulting but high-type borrowers so that
λ(1−p)pλ(1−p)+(1−λ)R − r0 > 0.) In this case, the outside bank also has the option of lending to defaulting
borrowers without requiring collateral.
The second-period profits of the inside bank i are given by
λpµi(pri2,N − r0)
1
S
∫ S
s∗N
+λ(1− p)µi(pri2,D)1
S
∫ S
s∗D
where ri2,N and ri2,D are the second-period interest rates for successful (non-defaulting) and defaulting
borrowers respectively, and s∗N and s∗D are the marginal switching costs.
If the outside bank j lends to defaulting borrowers without requiring collateral, its second-period
profits are given by
λpµi(pij2,N − r0)
1
S
∫ s∗N
0
ds+ λ(1− p)µi(pij2,S − r0)1
S
∫ s∗D
0
−(1− λ)r0,
where ij2,N is the interest rate charged to successful first-period borrowers of the inside bank, and ij2,S
the rate offered to defaulting ones.
If the outside bank requires collateral from defaulting borrowers, then the expression for second-period
profits is the same as under the assumption that defaulting borrowers are not creditworthy:
λpµi(pij2,N − r0)
1
S
∫ s∗N
0
ds+ λ(1− p)(1− µi)(pij2,D + (1− p)cj2(1− l)− r0)1
S
∫ s∗D
0
ds.
We present the banks’ second-period profits, first-period interest rates and intertemporal profits below.
11
Theorem 7. Second-period profits under the contract with and without collateral are given by, respec-
tively, Πi,c2 = µiλpS
49 + µi
λ(1−p)9S (2S + (1 − p)l)2 + (1 − µi)
λpS9 + (1 − µi)
λ(1−p)9S (S − (1 − p)l)2, and
Πi2 = µiλ
49S + (1− µi)
(λ9S − (1− λ)S
), where i = A,B.
First-period interest rates are given by ri,c1 = 1λp
(r0−λS3 + 2λ(1−p)2l
3
)and ri1 = 1
λp
(r0−λS3 +(1−λ)r0
),
respectively.
Intertemporal profits under the contract with and without collateral are given by, respectively, Πi,c =
λpS9 + λ(1−p)
9S (S − (1− p)l)2, and Πi = λS9 − r0(1− λ).
Proof. See Appendix.
We then summarize the decision to require collateral from defaulting borrowers.
Theorem 8. Banks prefer to offer contracts with collateral under information sharing and without collat-
eral in the absence of information sharing whenever (1−p)l9S (2S−(1−p)l) > r0(1−λ)
λ > (1−p)2l9S (2S−(1−p)l).
If there the cost of adverse selection is very high ( r0(1−λ)λ > (1−p)l9S (2S − (1− p)l)), then collateral is re-
quired under both information sharing regimes; when they are very low ( r0(1−λ)λ < (1−p)2l9S (2S− (1−p)l)),
then there are no collateral requirements under either regime.
Proof. The result is obvious when second-period poaching profits are compared.
Intuitively, when liquidation costs are so high that offering collateral contracts to all borrowers is
not justified to eliminate adverse selection, information sharing (even though it is partially informative)
may justify use of costly collateral by allowing to use it for part of the population based on repayment
history. When the cost of adverse selection is very low compared to the liquidation costs entailed by
collateral requirements, no collateral will be required under either regime; when the liquidation costs
are very low, then collateral will be required under both regimes. However, in the intermediate case,
collateral requirements will be observed under information sharing, but not in its absence.
4.3. Endogenous information sharing
So far we have taken the information sharing regime as given. In many countries, the existence of
credit registries is the result of decisions made by governments and/or central banks, and thus at least
to some extent exogenous to commercial bank decisions. However, we also frequently observe voluntary
information sharing, and it may be interesting to see when banks choose to do that given the potential
for collateral requirements.
Theorem 9. If adverse selection is very low ( r0(1−λ)λ < (1−p)2l9S (2S−(1−p)l)), then banks are indifferent
between the two information sharing regimes if defaulting borrowers are creditworthy, and do not share
information if they are not. If adverse selection is very high ( r0(1−λ)λ > (1−p)l9S (2S− (1− p)l)), then banks
prefer to share information provided that switching costs are high enough (S > (1−p)l2 ). In the intermediate
case ( (1−p)l9S (2S − (1− p)l) > r0(1−λ)
λ > (1−p)2l9S (2S − (1− p)l)), banks prefer to share information.
Proof. We get the decision rules by comparing intertemporal profits.
12
Intuitively, when adverse selection issues are low and defaulting borrowers are creditworthy, banks will
not try to separate borrowers, and they will have the same profits under both regimes. This is because
under both information regimes they try to poach among all first-period borrowers of the competing
banks without requiring collateral. If defaulting borrowers are not creditworthy, then under information
sharing banks do not try to attract their competitors’ defaulting borrowers, and given relatively low
adverse selection this is worse than not sharing information and bidding for the entire pool of borrowers.
When adverse selection issues are important and liquidation costs are low, banks will choose to use
collateral under both regimes. Profits will be higher in that case under information sharing if it is diffi-
cult for borrowers to switch, since increased switching by successful borrowers may otherwise reduce bank
profits. Finally, in the intermediate case, banks will prefer information sharing and require collateral only
from defaulting borrowers.
As in Gehrig and Stenbacka Gehrig and Stenbacka [24], if we have more than two banks and there is
competition for poaching the other banks’ first-period borrowers, information sharing becomes an issue
of indifference. However, it can be shown that our main result - a higher incidence of collateral under
information sharing - survives in that case.
4.4. Welfare
We have seen in the previous subsection that under moderate adverse selection, the use of collateral
will be observed under information sharing, but not in the absence of a credit bureau. Moreover, if banks
are the ones making the choice about sharing information, they will prefer to do so, since that brings
them higher profits. We can check whether this choice is associated with higher or lower overall welfare,
and see how the surplus is shared between banks and borrowers.
Theorem 10. 1. In the presence of moderate adverse selection ( (1−p)l9S (2S − (1 − p)l) > r0(1−λ)
λ >
(1−p)2l9S (2S − (1 − p)l)), welfare is higher under information sharing (when collateral is required),
then in its absence (when collateral is not required).
2. In the presence of moderate adverse selection ( (1−p)l9S (2S−(1−p)l) > r0(1−λ)
λ > (1−p)2l9S (2S−(1−p)l)),
the surplus of high-type borrowers is higher under information sharing than in its absence if cS <
3−8L(1−p)(8−6L) .
Although information sharing increases welfare in this case (and will be preferred by banks if they
have the possibility to choose), creditworthy borrowers are not necessarily better off. This is more likely
if switching costs are higher, liquidation costs are lower, and the private benefits of low-type borrowers
are lower. The less favorable conditions for high-quality borrowers can be seen as a disadvantage per se;
moreover, it indicates that marginally worthwhile projects may be abandoned as a result of frictions such
as the liquidation cost of collateral5.
5Total bank profits will of course be zero over the two periods. However, in the second period, the collateral-basedcontract will be preferred under information sharing. The proof is available upon request.
13
5. Conclusion
Banks have several alternative instruments that can be used to reduce adverse selection costs. We
examine two such devices in our paper: collateral and information sharing. While both serve the basic
purpose of selecting borrowers, we show that they may be complements rather than substitutes: infor-
mation sharing makes it more likely that collateral is required from borrowers with poor credit histories.
Given the cost of pledging collateral, and indeed its unavailability to many borrowers, our result points
to a potential downside of information sharing. This can be therefore added to the list on which we already
have the high initial interest rates (Gehrig and Stenbacka [24]) and the potential for “excessive” memory
(Vercammen [46]).
Our finding about the high use of collateral can be related to the results in Manove, Padilla and
Pagano [36], where they show that if the average quality of a borrower pool is too low collateral may
replace screening. We believe the interaction between borrower selection mechanisms is a promising area
for future research.
14
Acknowledgements
The views expressed in this article are our own and do not necessarily reflect the views of Norges Bank.
We would like to thank seminar participants at Norges Bank, NHH and Carlos III Business Department
for helpful comments and suggestions. All remaining errors are ours.
References
[1] Avery, R. B. , Bostic, R. W., Samolyk, K. A., 1998. The role of personal wealth in small business
finance. Journal of Banking and Finance 22, 1019-1061.
[2] Benmelech, E., Bergman, N. K., 2009. Collateral pricing, Journal of Financial Economics 91, 339360.
[3] Berger, A. N., Udell, G. F., 1995. Relationship Lending and Lines of Credit in Small Firm Finance.
Journal of Business 68, 351-381.
[4] Berger, A.N., Frame W.S., Ioannidou V., 2011. Tests of ex ante versus ex post theories of collateral
using private and public information. Journal of Financial Economics 1, 85-97.
[5] Berger, A. N., Espinosa-Vega, M. A., Frame, W. S., Miller, N. H., 2011. Why do borrowers pledge
collateral? New empirical evidence on the role of asymmetric information. Journal of Financial
Intermediation 20, 5570.
[6] Bernanke, B., Gertler, M., 1989. Agency Costs, Net Worth, and Business Fluctuations. American
Economic Review 79, 14-31.
[7] Bernanke, B., Gertler, M., Gilchrist, S., 1996. The Financial Accelerator and the Flight to Quality.
Review of Economics and Statistics 78, 1-15.
[8] Besanko, D., Thakor, A.V., 1987. Collateral and rationing: sorting equilibria in monopolistic and
competitive credit markets. International Economic Review 28, 671689
[9] Bester, H., 1985. Screening vs. Rationing in Credit Markets with Imperfect Information. The Amer-
ican Economc Review 75, 850-855.
[10] Bharath, S. T. , Dahiya, S. , Saunders, A., 2011. Lending Relationships and Loan Contract Terms.
Review of Financial Studies 24, 1141-1203.
[11] Bodenhorn, H., 2003. Short-Term Loans and Long-Term Relationships: Relationship Lending in
Early America. Journal of Money, Credit and Banking 35 485-505.
15
[12] Boot, A. W. A., Thakor, A. W., 1994. Moral Hazard and and Secured Lending in an Infinitely
Repeated Credit Market Game. International Economic Review 35, 899-920.
[13] Booth, J. R., Booth, L. C., 2006. Loan Collateral Decisions and Corporate Borrowing Costs. Journal
of Money, Credit and Banking 38, 67-90.
[14] Bouckaert, J., Degryse, H., 2006. Entry and Strategic Information Display in Credit Markets.
Economic Journal 116, 702-720.
[15] Brown, M., Jappelli, T., Pagano, M. 2009. Information Sharing and Credit Market Performance:
Firm-Level Evidence from Transition Countries. Journal of Financial Intermediation 18, 151-172.
[16] Cerqueiro G., Ongena S. and Roszbach K., 2012. Collateralization, Bank Loan Rates and Monitor-
ing: Evidence from a Natural Experiment. Unpublished working paper, Sveriges Riksbank.
[17] Chakraborty, A., Hu, C., 2006. Lending Relationships in Line-of-Credit and Nonline-of-Credit
Loans: Evidence from Collateral use in Small Business. Journal of Financial Intermediation 15,
86- 107.
[18] Chan, Y., Kanatas, G., 1985. Asymmetric Valuations and the Role of Collateral in Loan Agree-
ments. Journal of Money, Credit and Banking 17, 84-95.
[19] Chen, Y., 2006. Collateral, loan guarantees, and the lenders incentives to resolve financial distress.
Quarterly Review of Economics and Finance 46, 115.
[20] DellAriccia, G., Marquez, R., 2004. Information and bank credit allocation. Journal of Financial
Economics 72, 185214.
[21] Doblas-Madrid, A., Minetti R., Sharing information in the credit market: Contract-level evidence
from U.S.firms, Journal of Financial Economics, forthcoming.
[22] Djankov, S., McLiesh, C., Shleifer, A., 2007. Private credit in 129 countries. Journal of Financial
Economics 84, 99-329.
[23] Gan, J., 2007. Collateral, debt capacity, and corporate investment: Evidence from a natural exper-
iment. Journal of Financial Economics 85, 709734.
[24] Gehrig, T., Stenbacka, R., 2007. Information Sharing and Lending Market Competition with Switch-
ing Costs and Poaching. European Economic Review 51, 77-99.
[25] Gorton, G., Kahn, J., 2000. The Design of Bank Loan Contracts, Collateral, and Renegotiation.
Review of Financial Studies 13, 331-364.
[26] Hale, G., Santos, A., 2009. Do banks price their informational monopoly? Journal of Financial
Economics 93, 185206.
16
[27] Hertzberg, A., Liberti J., Paravisini D., 2011. Public Information and Coordination: Evidence from
a Credit Registry Expansion. Journal of Finance 66, 379-412.
[28] Holmstrom, B., Tirole, J., 1997. Financial intermediation, loanable funds, and the real sector.
Quarterly Journal of Economics 112, 663-691.
[29] Houston, J. F., Lin, C., Lin, P., Ma, Y., 2010. Creditor rights, information sharing, and bank risk
taking. Journal of Financial Economics 96, 485512.
[30] Inderst, R., Mueller, H. M., 2007. A lender-based theory of collateral, Journal of Financial Eco-
nomics 84, 826859.
[31] Jappelli, T., Pagano M. 1993. Information Sharing in Credit Markets. Journal of Finance 43, 1693-
1718.
[32] Jimenez, G., Salas, V., Saurina J., 2006. Determinants of collateral. Journal of Financial Economics
81, 255281.
[33] Jimenez, G., Salas-Fumas, V., Saurina, J., 2011. The Effects of Formal and Informal Contracting
in Credit Availability. Journal of Money, Credit and Banking 43, 2011.
[34] Kim, M., Kliger D., Vale B., 2003. Estimating Switching Costs and Oligopolistic Behavior. Journal
of Financial Intermediation 12, 25-56.
[35] Kiyotaki, N., Moore, J., 1997. Credit Cycles. Journal of Political Economy 105 211-248.
[36] Manove, M., Padilla, J. and Pagano M., 2001. Collateral versus Project Screening: A Model of
Lazy Banks. RAND Journal of Economics 32, 726-44.
[37] Miller, M., 2003. Credit Reporting around the Globe. In Credit Reporting Systems and the Inter-
national Economy, MIT Press, Cambridge, Massachusetts.
[38] Padilla, J., Pagano M. 1997. Endogenous Communication among Lenders and Entrepreneurial
Incentives. Review of Financial Studies 10, 205-236.
[39] Padilla, J., and Pagano M., 2000. Sharing Default Information as a Borrower Discipline Device.
European Economic Review 44, 1951-1980.
[40] Petersen, M., (2004) Information: Hard and Soft, Working Paper, Northwestern University.
[41] Rajan, R., Winton, A., 1995. Covenants and Collateral as Incentives to Monitor. Journal of Finance
50, 1113-1146.
[42] Sharpe, S., 1990. Asymmetric Information, Bank Lending, and Implicit Contracts: A Stylized
Model of Customer Relationships. Journal of Finance 45, 1069-1087.
17
[43] Stiglitz, J.E., Weiss, A., 1981. Credit Rationing in Markets with Imperfect Information, American
Economic Review 71, 1981, 393-410.
[44] Shy, O., 2002. A Quick-and-Easy Method for Estimating Switching Costs. International Journal of
Industrial Organization 20, 71-87.
[45] Stango, V., 2002. Pricing with Consumer Switching Costs: Evidence from the Credit Card Market.
The Journal of Industrial Economics 50, 475-492.
[46] Vercammen, J., 1995. Credit Bureau Policy and Sustainable Reputation Effects in Credit Markets,
Economica 62, 461-478.
18
6. Appendix
Proof of lemma 1
Consider the segment of borrowers that borrowed from bank i (i = A,B)in period 1. This population
has a mass of µi.
Suppose first that the outside bank j does not use collateral. The inside bank chooses an interest rate
ri2 for its (high-type) borrowers, and the outside bank chooses an interest rate ij2 for the inside bank’s
first-period customers. High-type borrowers with relatively low switching costs (s∗ < pri2 − pij2) will
switch to the outside bank. Low-type borrowers will also switch to the outside bank, since the outside
bank has no information that would allow it to reject them.
The inside bank will choose its interest rate to maximize its profits on first-period borrowers:
maxri2
µiλ(pri2 − r0)1
S
∫ S
s∗ds
The outside interest rate ij2 will be chosen to maximize profits on poached borrowers:
maxij2
µiλ(pij2 − r0)1
S
∫ pri2−pij2
0
ds− roµi(1− λ).
The first-order conditions are:
ri2 =pij2 + r0 + S
2p
ij2 =pri2 + r0
2p
Solving, we get ri2 = 3r0+2S3p and ij2 = 3r0+S
3p .
Suppose next that the outside bank decides to require collateral. In that case it will lend at the
interest rate ij2 to high-type borrowers with low switching costs (s < s∗ = pri2−pij2− c
j2), and will receive
the value of the collateral pledged by those borrowers, net of liquidation costs, in case of failure:
µiλ(pij2 + (1− p)cj2 − r0 − (1− p)l) 1
S
∫ pri2−pij2−(1−p)cj2
0
ds
Borrowers will be required to pledge an amount of collateral equal to their private benefit c. This is the
minimal amount that allows the separation of high- and low-type borrowers; the banks will not require
more than that given the cost of liquidating collateral. By assumption, borrowers have the required
collateral (their assets in place C are higher than c).
The incumbent bank lends at the interest rate ri2 to high-type borrowers with high switching costs,
and gets profits equal to µiλ(pri2 − r0) 1S
∫ Spri2−pi
j2−(1−p)cj2
ds.
19
The first-order conditions are:
ri2 =pij2 + (1− p)cj2 + r0 + S
2p
ij2 =pri2 − 2(1− p)cj2 + (1− p)l + r0
2p
Solving, we get ri2 = 3r0+2S+(1−p)l3p and ij2 = 3r0+S+2(1−p)l−3(1−p)c
3p .
Proof of Proposition 2
If collateral is not used for poaching, then total second-period profits for bank i are equal to:
Π2 = µiλ(pri2 − r0)1
S
∫ S
s∗ds+ µjλ(pii2 − r0)
1
S
∫ prj2−pii2
0
ds− roµj(1− λ).
Using the expressions for equilibrium interest rates derived in the previous lemma, we get that the
second-period profits can be written as:
Π2 = µiλ4
9S + (1− µi)
(λ
1
9S − (1− λ)r0
).
If collateral is required from borrowers that want to switch to the outside bank, then second-period
profits are given by:
Π2 = µiλ(pri2 − r0)1
S
∫ S
s∗ds+ µjλ(pii2 + (1− p)ci2 − r0)
1
S
∫ prj2−pii2
0
ds.
Using the expressions for equilibrium interest rates derived in the previous lemma, we get that the
second-period profits can be written as:
Π2 = µiλ1
9S
(2S + (1− p)l
)2+ (1− µi)λ
1
9
(S − (1− p)l
)2.
Proof of Proposition 4
The intertemporal profits are given by the sum of the profits over two periods: Πi = Π1i + Π2
i , where
Π1i = µi(λpr
i1−r0) and the equilibrium second-period profits are characterized above. These total profits
are linear in the market share µi and are expressed (for instance in the case where collateral required in
the second period) by
µi
(λpri1 − r0 + λ 4
9S
(S + (1− p)l
)2 − λ 19
(S − (1− p)l
)2)+ λ 1
9
(S − (1− p)l
)2The first term in brackets expresses profits resulting from the market share of that bank i acquires in
period 1. If those profits are positive, the other bank can charge a slightly lower first-period interest rate
r1 and take over the entire market. Therefore initial competition will drive these profits to zero, so that
for instance ri1 = 1λp
(r0 − λ 4
9S
(S + (1− p)l
)2 − λ 19
(S − (1− p)l
)2)when collateral is used in the second
period.
20
Proof of Proposition 6
Suppose first that the outside bank j does not use collateral. In that case, it does not lend to de-
faulting borrowers and chooses an interest rate ij2,N to maximize its second-period profits from successful,
switching borrowers λpµi(pij2,N − r0) 1
S
∫ s∗N0
, where s∗N = pri2,N − pij2,N . The inside bank i will charge
the maximal rate R for loans to high-type, defaulting borrowers, and will choose an interest rate for
defaulting borrowers ri2,N to maximize its profits on successful borrowers λpµi(pri2,N − r0) 1
S
∫ Ss∗N
. Solving
the two profit maximization problems, we get ri2,N = 3r0+2S3p and ij2,N = 3r0+S
3p , with a marginal switching
cost s∗N = S3 . Entering these interest rates in the expressions for inside and outside profits we get the
total second-period profits.
Next, if the outside bank uses collateral to select switching, defaulting borrowers, it will choose an
interest rate ij2,D and collateral to maximize profits on those borrowers λ(1 − p)(1 − µi)(pij2,D + (1 −
p)cj2(1− l)− r0) 1S
∫ s∗D0
ds, where s∗D = pri2,D − pij2,D − (1− p)cj2. It will also choose an interest rate ij2,N
to maximize profits on successful switching borrowers as shown above. The inside bank will choose an
interest rate ri2,D to maximize its profits on high-type, defaulting borrowers λ(1 − p)µi(pR − r0) 1S
∫ Ss∗D
,
and an interest rate ri2,N to maximize profits on successful borrowers, as above. Solving, we get that
ij2,D = 3r0+S−(1−p)(3c−2l)3p and ri2,D = 3r0+2S+(1−p)l
3p , with s∗D = S3 + (1− p)c− 1
3 (1− p)l. Entering these
interest rates in the expressions for inside and outside profits we get the total second-period profits.
The first-period interest rate r1 will set the two-period profits from acquiring a market share µi to
zero. If collateral is not used for poaching in the second period, we have that µi
(λpr1 − r0 + λpS3 +
λ(1− p)(pR − r0))
= 0, thus the first-period interest rate is r1 = 1λp
(r0 − λpS
3 − λ(1− p)(pR − r0))
. If
collateral is used, µi
(λpr1 − r0 + 1
3λpS + λ(1−p)9S (2S + (1− p)l)2 − λ(1−p)
9S (S − (1− p)l)2)
= 0, therefore
r1 = 1λp
(r0 − λS
3 −2λ(1−p)2l
3
).
Proof of proposition 7
If defaulting borrowers are creditworthy, they will receive a loan from the outside bank even in the
absence of collateral. In that case, the inside bank i chooses an interest rate ri2,N to maximize profits from
lending to successful borrowers λpµi(pri2,N − r0) 1
S
∫ Ss∗N
, and an interest rate ri2,D to maximize its profits
accruing from loans to defaulting high-type borrowers λ(1− p)µi(pri2,D) 1S
∫ Ss∗D
. The outside bank chooses
an interest rate ij2,N to maximize profits from switching successful borrowers λpµi(pij2,N−r0) 1
S
∫ s∗N0
ds and
a rate ij2,D to maximize profits on switching defaulting borrowers λ(1− p)µi(pij2,S − r0) 1S
∫ s∗D0−(1−λ)r0.
The cutoff switching costs are s∗N = pri2,N − pij2,N and s∗D = pri2,D − pi
j2,D respectively. Solving, we get
ri2,N = ri2,D = 3r0+2S3p and ij2,N = ij2,D = 3r0+S
3p . Using these expressions we get the second-period profits
stated in the proposition. The first-period interest rate r1 can be calculated as in the proof of Proposition
6.Intertemporal profits are therefore λS9 − (1− λ)r0.
When collateral is used, then expression for bank profits will be the same as in the case where
defaulting borrowers are not creditworthy.
Proof of proposition 10
21
If (1−p)l9S (2S−(1−p)l) > r0(1−λ)
λ > (1−p)2l9S (2S−(1−p)l), collateral will be required under information
sharing and it would not be required under information sharing.
Under information sharing, overall welfare is
W = λ(pR− r0)− λ(1− p)S−2(1−p)l
3
Sl − λ(1− p)
S
∫ S−2(1−p)l3
0
sds− λp
S
∫ S3
0
sds,
that is, the surplus generated by lending to high-type borrowers, less the liquidation costs if collateral in
case of default, less the switching costs of high-type borrowers.
Without information sharing, overall welfare is
W = λ(pR− r0)− (1− λ)r0 −λ
S
∫ S3
0
sds− (1− λ)
S
∫ S
0
sdS,
that is, the surplus generated by lending to high-type borrowers, less losses from lending to low-type
borrowers, less the switching costs of both high- and low-type borrowers.
Welfare is higher under information sharing if
1− λλ
r0 > (1− p) (2S(1− p)l − (1− p)2l2
9S− (1− p)2l(S + 3(1− p)l)
9S− (1− λ)
S
2.
This is a weaker condition than the condition for this case:
1− λλ
r0 > (1− p) (2S(1− p)l − (1− p)2l2
9S,
therefore welfare is always higher under information sharing.
In the same case, when collateral is required under information sharing and not without information
sharing, we have a higher surplus for high-type borrowers under information sharing if
2
3p2(R− 3r0 + 2S
3p
)+
1
3p2(R− 3r0 + S
3p
)− p S
18+ p(1− p)
(R− 3r0 + 2S + (1− p)l
3p
)S − S−2(1−p)l3
S
+ p(1− p)(R− 3r0 + S + 2(1− p)l − 3(1− p)c
3p
)S − 2(1− p)l3S
− (1− p) (s− 2(1− p)l)2
18S− (1− p)2l s− 2(1− p)l
3S
>2
3p(R− 3r0 + 2S
3p
)+
1
3p(R− 3r0 + S
3p
)− S
18.
This condition can be rewritten as
c
S<
3− 8L
(1− p)(8− 6L).
22